OTTAWA — The artificial intelligence race is only just getting started, but tech giants can’t be allowed to block competition, European antitrust chief Margrethe Vestager says.

The Danish policymaker will finish her decade-long stint at the European Commission at the end of next month. Vestager has frequently gone after Silicon Valley’s biggest firms, launching landmark investigations into Apple, Amazon, Facebook and Google. She also spearheaded the EU’s Digital Markets Act, designed to force large tech firms to open up their app stores, payment systems and data to other developers. 

Antitrust regulators in other countries have followed suit. In the U.S., the Biden administration has more recently launched several cases against tech giants, while Canada’s Liberal government has moved to update competition laws to address what critics call unfair business practices.

Vestager, the European Commission’s executive vice-president for a Europe fit for the digital age, is midway through a three-day visit to Canada. On Wednesday, she met with Canadian Innovation Minister François-Philippe Champagne in Montreal, where the two also talked antitrust and AI policy at a public event. 

In an interview with The Logic in Ottawa on Thursday, Vestager discussed what policymakers can do to ensure the AI market remains open and that the technology is used appropriately, as well as why governments need to align on major digital issues.

The interview has been edited for length and clarity. 

Do jurisdictions like the EU and Canada need continental or national champions to compete with the U.S. and China in areas like artificial intelligence or electric vehicle batteries?

I definitely think that you need actors in these markets. I’m more doubtful if you need to compete on size. You need to compete on innovation, on getting to the customers.

It is important that you find manufacturing capacity on these essential technologies in every jurisdiction. We learned it the hard way on pipeline gas—the dependency on Russia has been really negative on Europe and the war on Ukraine. It is important to keep importing, but also to be able to both export and produce for your own supply.

What about in intangible markets like software?

I think the same applies. If you look at what we’re trying to do on AI, we are in process to enable small- and medium-sized companies to use the [EU’s] network of hyperscale computers to develop artificial intelligence in Europe. These companies would not necessarily have the access to this kind of compute, so if we can have that kind of AI factory, that would be really good for us. 

That doesn’t mean that AI developed in other jurisdictions will not be used in Europe. But it means that we will have more to offer, because the development of AI is only beginning.

OpenAI seems the most likely company to join the list of tech giants. Whichever that next company ends up being, it seems inevitable that it’ll come out of Silicon Valley, or at least the U.S. Why is it that Europe, Canada or other jurisdictions haven’t been able to create companies of that importance, if not size?

I think the approach is a bit different. One of the things that we’re doing with the [EU] AI Act is to create markets where it is absolutely essential that you can trust the technology that you use. It is essential that technology serves humans, and that is our guiding principle. The public sectors in Europe [are] big. There are a lot of positive-use cases for AI, if you can trust the machine not to discriminate among citizens; if that was a risk, then governments, municipalities and regions would scare off. 

We see this as market-creating, and we also see smaller European companies moving into that market. It is still nascent, but I take it for granted that it will come.

Some U.S. companies are withholding AI services from Europe—Meta’s latest Llama model and parts of OpenAI’s latest model are not available. Does that disadvantage companies in Europe who can’t access that because of the regulation that you have? 

It must be very thought-provoking in the jurisdictions where these products are launched. In Europe, regulation is, I think, quite straightforward. You need to protect customer data. You cannot do whatever with the data that you gobble up; you have obligations. 

I would be more concerned if I was in one of the jurisdictions where there’s nothing to protect your data or make sure that there are guardrails for the algorithmic technology to serve you. If you want to leave a market for someone else, well, then it is left for someone else. 

In the AI space, cloud providers are investing in startups making foundation models that are then using their cloud services. In general, are those kinds of deals a worry for you?

It very much depends on the specific deal. What could be worrying for us is if there’s no check on what Big Tech is doing in terms of embedding AI companies in their ecosystem, making that market even more of a walled garden. For innovation reasons, for consumer choice, we would want those markets to be open and contestable. That’s the very reason that we have the Digital Markets Act. 

You were the first major-economy regulator with this new approach of focusing more on digital markets. How would you grade Canada’s response in comparison to the work you’ve been doing and the U.S. has been doing?

I was with [Innovation] Minister [François-Philippe] Champagne yesterday, and I think he’s rightly very proud of the changes to Canadian competition legislation to give it much stronger teeth. It’s important for a competition agency to have that kind of independence, resources and tools in their toolbox to do that job. 

Of course, it’s much more difficult for a smaller jurisdiction than it is for a larger jurisdiction. But it’s important for us that Canada is also part of the effort, because the more we are acting on this, the better.

In Montreal yesterday, you talked about the need to come together and pass domestic laws to deal with global phenomena like competition and AI. Even if economies the size of the EU and Canada are taking similar approaches, does it matter unless the U.S. stays aligned?

It’s really important that the U.S. is aligned. It was one of the many successes of the [EU-U.S.] Trade and Technology Council that we agreed on the approach to AI very early on—to look at use cases and to look at what is most risky. You see that in the U.S., translated into the work of the [Biden administration] and companies signing up voluntarily [to AI guardrails]. 

We can get somewhere with voluntary commitments, but having legislation that has bite is what makes voluntary commitments much more likely to be fulfilled. In my view it can be done in a way that is market-creating, because you can trust technology while [being] hands off when it comes to innovation and research. That is the important enforcement message for the next commission. 

You mentioned the AI Factories program earlier. Canada, the U.K., and others are trying to set up their own sovereign compute capacity. Do governments risk driving up the cost for resources like chips or skilled talent if everyone’s trying to build AI compute right now?

We are giving access to existing compute. If you want to have something more specific to AI, it would take a year-and-a-half to establish it, so the first thing is to make sure that existing capacity can be used [for] what is high priority. There are limitations of access to GPUs, [so] it’s important that the resources we have are being used to its fullest. 

What is important for us is that the workforce is developing at the same speed. I found it very inspiring to visit Singapore some months ago. They have a program called AI Apprenticeships, where they take in people with a varied set of backgrounds to work with specific businesses on AI deployment. In the process, everybody learns. 

Those are very interesting models.They have this double-sided effect: they help the business that wants to incorporate technology, and they develop the workforce at the same time. We have similar programs, but I was really taken with this because it was so hands-on. Many smaller businesses think, “Oh my god. How do I approach this?” 

We can learn a lot from one another, in order not to fall in that trap [where] everyone wants to do the same thing for the same very limited resources available. 

Have you had any discussions with Minister Champagne about sharing compute resources?

No, we have not. We share a lot, and it has been such a pleasure to work with him within the G7. It was so precious that the Japanese took the leadership on [an] AI code of conduct, and with Minister Champagne we really actively pushed in order to enable this to happen. 

The EU and Canada have a lot in common in what we want to achieve. As I said yesterday when we were opening the MonarQ, the Canadian-made quantum computer, that’s the kind of partner you want: someone who can pull this off all by themselves. You can say, “Oh, but there’s no little European flag.” That is not the point. The point is that these people are great and we want to work with them.